[Python-Dev] stats.py (was 'summing a bunch of numbers ')

Chad Netzer cnetzer@mail.arc.nasa.gov
21 Apr 2003 15:07:28 -0700

On Mon, 2003-04-21 at 05:30, Guido van Rossum wrote:
> Since it already exists as a 3rd party package, we should definitely
> not try to duplicate the effort.  Then the question is, is it enough
> to point to the 3rd party package or does it deserve to be
> incorporated into the core?  We can't go and incorporate every useful
> 3rd party package into the core

True.  I just happen to be of the opinion that a statistics package is
the single most practical and useful mathematics pakage that can be
added to a language, after basic linear algebra (which isn't in the
core... Hmmm)

> OTOH, having it in the core, with decent documentation, might prevent
> naive wannabe-statisticians like myself from misremembering how
> standard deviation is implemented, or when to use it. :-)

My concern, like yours, is that this kind of thing is probably
reimplemented a LOT (at least the simple stats functions).  If we
adopoted it, I would actually favor keeping it fairly lightweight
(although t-tests and even ANOVA should go in).  Heavyweight users could
always download a separate add on package.

Of course, the stats.py package (and it's SciPy cousin) DOES seem to be
well maintained, so perhaps the issue is just making sure those that
might need it can easily download and install it (ie. promote it to
distributions, give it proper promotion on Vaults of Parnassus, mirror
it, etc.).

My personal preference would be make it standard (as I propably would
like NumPy to become).  I like to use it in my unittests. :)  That may
not be the consensus, though.


Chad Netzer
(any opinion expressed is my own and not NASA's or my employer's)